For radio frequency identification using the LANDMARC indoor positioning system,k-nearest neighbor reference tags that meet certain conditionswere selected,and the final position of the unknown tag was estimated according to those reference tags.However,the methods of selecting of the number of adjacent reference tags are also blindness.To solve this problem,the reference tag thatis nearest the unknown node is selected as an unknown point; then,the optimal number k is obtained by searching all feasible k values.Simulation results demonstrate that the performance of the improved LANDMARC algorithm is better than the traditional LANDMARC algorithm,and the accuracy improved by approximately 10%.